Article title: How do institutional factors shape PhD completion rates? An analysis of long-term changes in a European doctoral program
Authors: J. Skopek, M. Triventi, H.-P. Blossfeld
Journal: Studies in Higher Education
DOI: https://doi.org/10.1080/03075079.2020.1744125
The article was originally published with wrong figure sequences. The correct figures are listed as below,
Figure 2. Entries to the PhD program from 1976 to 2012. Source: own calculations based on archival records.
![Figure 2. Entries to the PhD program from 1976 to 2012. Source: own calculations based on archival records.](/cms/asset/e2255212-6334-4d51-80d2-1d64ba0da64f/cshe_a_1754697_f0002_ob.jpg)
Figure 3. Probability of completion by entry cohort. Note: Proportion of admitted students who submitted their thesis up to a certain time. Estimated via non-parametric Kaplan–Meier estimation for each cohort.
![Figure 3. Probability of completion by entry cohort. Note: Proportion of admitted students who submitted their thesis up to a certain time. Estimated via non-parametric Kaplan–Meier estimation for each cohort.](/cms/asset/e24e0980-075d-43db-a24d-056f9125746a/cshe_a_1754697_f0003_ob.jpg)
Figure 4. Time evolution of hazard rates of thesis submission for three student cohorts. Note: Smoothed hazard rates by cohort using Nelson–Aalen estimation. Hazard rate on yearly scale.
![Figure 4. Time evolution of hazard rates of thesis submission for three student cohorts. Note: Smoothed hazard rates by cohort using Nelson–Aalen estimation. Hazard rate on yearly scale.](/cms/asset/2ae47e57-8286-4591-b6b4-906aa4069309/cshe_a_1754697_f0004_ob.jpg)
Figure 5. Changes of average completion hazard of entry cohorts shortly before and after the first reform in 1989 and the second reform in 2000. Note: Estimated for cohorts who entered immediately five years before and after the reforms in 1989 and 2000 using an exponential rate model. Hazard rate on yearly scale.
![Figure 5. Changes of average completion hazard of entry cohorts shortly before and after the first reform in 1989 and the second reform in 2000. Note: Estimated for cohorts who entered immediately five years before and after the reforms in 1989 and 2000 using an exponential rate model. Hazard rate on yearly scale.](/cms/asset/84050882-998e-411b-8615-fec0dae899d6/cshe_a_1754697_f0005_ob.jpg)
Figure 6. PhD completion by entry cohort unconditional (Model 4) and conditional on covariates (Model 5). Notes: Predictions of hazard rate (yearly) and cumulative probability based on Models M4 and M5. Covariate values fixed at sample mean.
![Figure 6. PhD completion by entry cohort unconditional (Model 4) and conditional on covariates (Model 5). Notes: Predictions of hazard rate (yearly) and cumulative probability based on Models M4 and M5. Covariate values fixed at sample mean.](/cms/asset/3c0146a4-9232-4729-9251-6ce96e54382f/cshe_a_1754697_f0006_ob.jpg)
Figure 7. Covariate effects on cumulative probability of PhD completion (Model 5). Notes: Difference in predicted probabilities of completion based on Model 5. Effects evaluated at the sample average. Non-significant effects in grey. range = denotes range effect going from minimum to maximum sample values (see Table 1).
![Figure 7. Covariate effects on cumulative probability of PhD completion (Model 5). Notes: Difference in predicted probabilities of completion based on Model 5. Effects evaluated at the sample average. Non-significant effects in grey. range = denotes range effect going from minimum to maximum sample values (see Table 1).](/cms/asset/e36351b6-f6ff-4297-b123-9a3e0386b20c/cshe_a_1754697_f0007_oc.jpg)
Figure A1. Smoothed non-parametric hazard estimates (dashed line) and hazard based on model prediction (solid line). Note: Non-parametric hazards estimated by Nelson–Aalen estimator with a kernel density smoothing (Gaussian kernel). Model predictions based on M0 in Table 2. Hazard rate on yearly scale.
![Figure A1. Smoothed non-parametric hazard estimates (dashed line) and hazard based on model prediction (solid line). Note: Non-parametric hazards estimated by Nelson–Aalen estimator with a kernel density smoothing (Gaussian kernel). Model predictions based on M0 in Table 2. Hazard rate on yearly scale.](/cms/asset/37042703-be13-494f-8d5f-0bf78d254873/cshe_a_1754697_f0008_ob.jpg)